als_nmf: Alternating-least-squares non-negative matrix approximation

In jonalim/mfBiclust: Pipelines for Bi-Clustering Using Matrix Factorization

Description

Approximates a non-negative matrix as the product of two non-negative matrix factors using the alternating-least-squares algorithm by Paatero and Tapper (1994).

Usage

als_nmf(A, k, reps = 4L, maxIter = 100L, eps_conv = 1e-04,
  verbose = FALSE, ...)

Arguments

A	the matrix to factorize
k	the number of factors to calculate
reps	the number of replications to choose from
maxIter	the maximum number of least-squares steps
eps_conv	convergence tolerance
verbose	Print the mean squared error every 10 iterations
...	Present for compatibility. als_nmf has no other parameters

Details

Factorization is performed reps times, then the result with the minimum mean squared-error is returned. als_nmf() is fast for a small number of biclusters, but running time rapidly increases with k.

Value

a genericFit-class object

jonalim/mfBiclust documentation built on May 4, 2019, 4:13 a.m.